BlockTree currently implements its own binary search tree keyed by block size. Because it is not self-balancing, insertion order can make it unbalanced, degrading performance.

We should replace the internal tree logic with the utilities intrusive red-black tree and route insert/remove through to that. This would make the tree balanced, while removing duplicate tree handling code. BlockTree would remain as the layer coordinating memory blocks; managing intrusive node storage within blocks and handling equal-sized block lists.